Interprocessor collective communication library (InterCom)

Barnett, Mike; Shuler, Lance; Geijn, Robert van de; Gupta, Saurabh; Payne, D. G.; Watts, Jerrell

doi:10.1109/shpcc.1994.296665

Cited by 102 publications

(59 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sanders et al [10], Sistare et al [11] and Tipparaju et al [12] discuss various approaches to optimize collective communication algorithms for SMP clusters. Some work focuses on using different algorithms for different message size, such as [13,14]. None of previous work shows how it interacts with existing process placement approaches which are based on point-to-point communications and may also result in sub-optimal mappings.…”

Section: Related Workmentioning

confidence: 99%

Process Mapping for MPI Collective Communications

Zhang

Zhai

Chen

et al. 2009

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract.It is an important problem to map virtual parallel processes to physical processors (or cores) in an optimized way to get scalable performance due to non-uniform communication cost in modern parallel computers. Existing work uses profile-guided approaches to optimize mapping schemes to minimize the cost of point-to-point communications automatically. However, these approaches cannot deal with collective communications and may get sub-optimal mappings for applications with collective communications.In this paper, we propose an approach called OPP (Optimized Process Placement) to handle collective communications which transforms collective communications into a series of point-to-point communication operations according to the implementation of collective communications in communication libraries. Then we can use existing approaches to find optimized mapping schemes which are optimized for both point-to-point and collective communications.We evaluated the performance of our approach with micro-benchmarks which include all MPI collective communications, NAS Parallel Benchmark suite and three other applications. Experimental results show that the optimized process placement generated by our approach can achieve significant speedup.

show abstract

Section: Related Workmentioning

confidence: 99%

Process Mapping for MPI Collective Communications

Zhang

Zhai

Chen

et al. 2009

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…For example, op will be an addition operator for a count aggregation and a maximum operator for a max aggregation. Such collective communication for computing a final global meta array can be implemented efficiently on most parallel computers and networks of workstations [1]. Thus, the overhead for combining global meta arrays is expected to be negligible because the volume of communication is only P words per processor.…”

Section: Parallel Bucket Algorithmmentioning

confidence: 99%

Efficient algorithms for large-scale temporal aggregation

Moon¹,

Lopez

Immanuel³

2003

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

Abstract-The ability to model time-varying natures is essential to many database applications such as data warehousing and mining. However, the temporal aspects provide many unique characteristics and challenges for query processing and optimization. Among the challenges is computing temporal aggregates, which is complicated by having to compute temporal grouping. In this paper, we introduce a variety of temporal aggregation algorithms that overcome major drawbacks of previous work. First, for small-scale aggregations, both the worst-case and average-case processing time have been improved significantly. Second, for large-scale aggregations, the proposed algorithms can deal with a database that is substantially larger than the size of available memory. Third, the parallel algorithm designed on a shared-nothing architecture achieves scalable performance by delivering nearly linear scale-up and speed-up, even at the presence of data skew. The contributions made in this paper are particularly important because the rate of increase in database size and response time requirements has out-paced advancements in processor and mass storage technology.Index Terms-Temporal databases, temporal aggregation, scalable query processing, data partitioning, balanced tree algorithm, merge-sort algorithm, temporal query processing, aggregate queries.

show abstract

“…To counter the effects of congestion, carefully designed topology aware algorithms have been used for collectives on such networks [4,5]. In addition, there is a set of topology oblivious algorithms which perform reasonably well on most systems [7,8,3]. However, most of these algorithms do not seem to be a good fit for two-tier direct networks as they may not be able to make full use of the all-to-all connectivity in two-tier direct networks.…”

Section: Introductionmentioning

confidence: 99%

Collectives on Two-Tier Direct Networks

Jain

Lau

Kalé

2012

Recent Advances in the Message Passing Interface

View full text Add to dashboard Cite

Abstract. Collectives are an important component of parallel programs, and have a significant impact on performance and scalability of an application. To obtain best performance, platform specific implementations of various parallel programming frameworks, such as MPI and Charm++, are done. As a result, when systems with new network topologies are built, new topology aware algorithms for collectives are added to these frameworks that also contain the topology oblivious algorithms. In this paper, we propose topology aware algorithms for collectives performed on two-tier direct networks such as IBM PERCS and Dragonfly. We observe that, for large message operations, significant performance gains can be made by taking advantage of large number of links in a twotier direct network. We evaluate proposed algorithms using an analytical model based on link utilization.

show abstract

Interprocessor collective communication library (InterCom)

Cited by 102 publications

References 3 publications

Process Mapping for MPI Collective Communications

Process Mapping for MPI Collective Communications

Efficient algorithms for large-scale temporal aggregation

Collectives on Two-Tier Direct Networks

Contact Info

Product

Resources

About